Modeling of Magnetization Processes of 3-D-Printed Fe–Si Components by Means of an Artificial Neural Network Implemented in a Finite-Element Scheme

Abstract

Presently, iron–silicon (Fe–Si) alloys are considered the optimal materials for the fabrication of magnetic cores for electric motors. Additive manufacturing (AM) facilitates the fabrication of Fe–Si alloys with elevated silicon concentrations, highly advantageous to limit the electric conductivity and maximize the magnetic permeability. Given the novelty of the approach, there is a paucity of research on hysteresis modeling and simulations of components fabricated by AM. In this article, the focus is on a Fe–Si 3.7% wt Si fabricated by AM. The hysteresis has been modeled by means of an artificial neural network (ANN) trained on the quasi-static (1 Hz) hysteresis loops measured using the volt-amperometric experimental setup on the bulk material, a full-section toroid. The trained ANN is subsequently implemented in a finite-element method (FEM) software to simulate the hysteresis on a material sample with air gaps and at higher frequencies never seen in the training phase. This work demonstrates, for the first time, the robust predictive capability of an ANN–FEM framework. A key contribution is the validation of the model under purely predictive conditions, using a geometry and frequency range entirely unseen during training. The simulated results have been compared with measurements and with results obtained with the classical Jiles–Atherton (JA) model. The correlation between the ANN results and the experimental data is substantial, consistent with the JA results, and in certain instances, superior.